AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

Learning Representation for Clustering Via Prototype Scattering and Positive Sampling

Zhizhong Huang, Jie Chen, Junping Zhang, Hongming Shan

TPAMI 2022 · Label-efficient and robust learning › Recognition and long-tail · approach: Correspondence and causal representation learning · 147 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add prototype scattering loss (PSL) (ablation)ProPos · CIFAR-10ACC: 87.9 -> 91.6Improved
add prototype scattering loss (PSL) (ablation)ProPos · CIFAR-20ACC: 55.0 -> 57.8Improved
add positive sampling alignment (PSA) (ablation)ProPos · CIFAR-10ACC: 90.3 -> 91.6Improved
add positive sampling alignment (PSA) (ablation)ProPos · CIFAR-20ACC: 55.1 -> 57.8Improved
add PSL uniformity term (ablation)ProPos · CIFAR-10ACC: 87.8 -> 91.6Improved
add PSL alignment term (ablation)ProPos · CIFAR-10ACC: 92.1 -> 91.6No clear effect
add PSL alignment term (ablation)ProPos · CIFAR-20ACC: 57.3 -> 57.8Improved
replace CC cluster head with PSL (ablation)CC · CIFAR-10ACC: 74.6 -> 83.4Improved
integrate CC loss into BYOL (ablation)BYOL · CIFAR-10ACC: 87.8 -> 86.3Hurt
integrate PCL loss into BYOL (ablation)BYOL · CIFAR-10ACC: 87.8 -> 85.3Hurt
add PSL and PSA to BYOL on long-tailed dataBYOL · CIFAR-20-LTACC: 34.6 -> 39.0Improved
add PSL on top of PSA on long-tailed data (ablation)ProPos w/o PSL · CIFAR-10-LTACC: 42.7 -> 43.9Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy